The Importance of Transaminases Flare in Liver Elastography: Characterization of the Probability of Liver Fibrosis Overestimation by Hepatitis C Virus-Induced Cytolysis

Background: Liver stiffness measurement (LSM) is crucial for appropriate fibrosis staging in patients with ongoing hepatitis C virus (HCV) infection. However, there is still an ongoing debate on the impact of serum transaminases (aspartate-aminotransferase, AST; alanine-aminotransferase, ALT) on LSM. Methods: We selected 110 patients undergoing HCV eradication therapy with LSM compatible with significant liver fibrosis. LSM was evaluated prior to therapy and one year after HCV eradication. Results: LSM showed a median decrease of 35% from baseline values, and 67 (61%) patients showed posttreatment values compatible with lower fibrosis stages. We developed two logistic regression models to determine the probability of liver fibrosis overestimation according to serum transaminase. The probability of overestimation of two or more fibrosis grade is equal to (1) 50% for AST of 99 IU/L (2.2 ULN) and ALT of 90.5 IU/L (2 ULN), (2) 80% for AST of 123.5 IU/L (2.74 ULN) and ALT of 101.5 IU/L (2.25 ULN), and (3) reaches 100% for AST of 211 IU/L (4.7 ULN) and ALT of 140 IU/L (3.1 ULN). Conclusions: This study highlights the impact of serum transaminases on LSM. We believe that our findings may serve as a reference point for appropriate fibrosis stratification by liver elastography in patients with HCV infection.


Introduction
Hepatitis C Virus (HCV) infection represents a significant cause of chronic liver disease, with approximately 70 million chronically infected individuals worldwide [1]. Due to the lack of spontaneous virus clearance, most of the patients (approximately 55%-85%) develop chronic HCV infection, which over a period of 20-30 years could lead to liver cirrhosis in 10%-40% of cases [2]. Fortunately, the advent of second-generation direct antiviral agents (DAAs) has changed the natural history of chronic HCV infection [3]. Assessment of liver disease progression (i.e., liver fibrosis staging) is of particular importance, as the choice of the treatment regimen, post-treatment prognosis, and follow-up depend on the stage of fibrosis [4]. According to European Guidelines, noninvasive techniques (NITs) should be preferred to liver biopsy to assess liver fibrosis before therapy initiation [4]. Among the different NITS, liver elastography has revolutionized-due to its safety and accuracy-the everyday clinical quantification of liver fibrosis [5]. However, elastography is burdened by several confounding factors (such as serum transaminase flares, cholestasis, hepatic venous congestion, right heart failure, and steatosis) that could lead to inaccurate staging of liver fibrosis [6]. In particular, preliminary studies reported that virus-induced necroinflammatory activity, in the form of alanine transaminases (ALT) flares with values greater than ten times the upper limit of normal (ULN), may lead to falsely increased liver stiffness measurement (LSM) [7]. More recent findings showed that even patients with lesser ALT increase (≥2 ULN) had higher LSM if compared to patients with normal ALT [8,9]. In addition, this hypothesis was further explored through a significant decrease of LSM in patients with hepatitis B virus (HBV) infection after 3months of antiviral treatment and ALT normalization [10]. Similar findings were found in a cohort of patients with HCV infection, in which patients with a higher histological degree of inflammation had higher LSM [11]. To definitively assess the influence of liver inflammation on liver elastography, we evaluated changes in LSM in a cohort of patients with HCV infection and LSM compatible with severe liver fibrosis one year after the end of treatment.

Material and Methods
The study was carried out following the guidelines of the local Ethics Committee for conducting research involving humans (ID: 2783 -approved the 21 st of May 2019). The study was conducted according to the criteria set by the declaration of Helsinki. Each subject signed informed consent before participating in the study.
The main endpoint of therapy was defined as a sustained viral response (SVR), equivalent to undetectable HCV-RNA (≤15 IU/mL) in serum or plasma 12 weeks (SVR12) or 24 weeks (SVR24) after the end of therapy [13]. Laboratory tests were repeated every four weeks during the treatment and three months after the end of therapy and at the one-year interval after SVR12.

Ultrasound and Elastography
Both ultrasound and elastography examinations were performed using Philips Affiniti 70 (Philips Healthcare, Netherlands) ultrasound machine. Prior to treatment and one year after SVR12, each patient underwent an ultrasonographic examination of the liver, gall bladder, spleen, and kidneys. In order to assess the severity of liver steatosis, we used the Hamaguchi Score (HS) [14], which evaluates hepatorenal echo contrast, liver brightness, deep attenuation of the echoes and vessel blurring. Portal vein Doppler examination and splenic measurement were available only one year after SVR12. Their acquisition methods were described elsewhere [15][16][17]. In addition, each patient underwent LSM (prior to treatment and one year after SVR12) and spleen stiffness measurement (SSM) (one year after SVR12), with point shear wave elastography (pSWE) using the ElastPQ evaluation protocol. Their acquisition methods and reliability characteristics are described elsewhere [18]. We used the following cut-off values: mild/no fibrosis (F0-1), 4.1-5.5 kPa; significant fibrosis (F2), 5.51-7.5 kPa; advanced fibrosis (F3), 7.51-11.00 kPa; liver cirrhosis (F4) >11.01 kPa [19,20].
We excluded patients with HS >3 and serum cholesterol >250 mg/dL. We also excluded patients with factors that could potentially influence splenoportal dynamics, including pregnant women, patients with current alcohol abuse, decompensating events (such as hepatic encephalopathy, variceal hemorrhage, ascites, and spontaneous bacterial peritonitis), previous endoscopic EVs banding ligation, ongoing intake of nonselective beta-blockers (NSBB), history of portal vein thrombosis, placement of transjugular intrahepatic portosystemic shunt (TIPS), noncirrhotic causes of PH, current/recent diagnosis of hepatocellular carcinoma [22], presence of collateral hepatofugal shuntings and signs of liver failure. We also excluded patients with heart failure and/or documented congestive hepatopathy.

Statistical Analysis
According to the size of our sample, the Shapiro-Wilk test was performed to verify the normal distribution of variables. Normally distributed variables were reported as mean (± standard deviation, SD), whereas other variables were reported as median (Quartile 1; Quartile 3). Differences between continuous variables were examined using the Student t-test (if normally distributed) or Mann-Whitney U test (if not-normally distributed). Patients were divided into four groups according to fibrosis downstaging: group 1 (G1), consisted of patients who had LSM compatible with F4 one year after SVR12; group 2 (G2), consisted of patients who had LSM compatible with F4 prior to treatment and F3 one year after SVR12; group 3 (G3), consisted of patients who had LSM compatible with F≥3 and F2 one year after SVR12; group 4 (G4), consisted of patients who had LSM compatible with F≥3 and F0-1 one year after SVR12. We did not employ any corrections for multiple comparisons.
To determine the collinearity between variables, we employed the variance inflation factor (VIF). Liver fibrosis overestimation probability was studied through logistic regression models [23]. In particular, the risk of overestimation of one stage of fibrosis (M1), was calculated using as dependent variable 0 = no (i.e., patient did not experience downstage of one fibrosis stage) and 1 = yes (patient experienced downstage of one fibrosis stage), and serum AST and ALT as independent variables. The risk of overestimation of two or more stages of fibrosis (M2), was also calculated using as dependent variable 0 = no (i.e., patient did not experience downstage of two or more fibrosis stages) and 1 = yes (patient experienced downstage of two or more fibrosis stages), and as independent variable serum AST and ALT. The independent variables were all modeled as continuous. In order to calculate the expected probability plot, each linear predictor (LP) was employed in the following function: We compared models using the Akaike information criterion (AIC) and the Bayesian information criterion (BIC) and calculated Nagelkerke pseudo-R 2 and the area under the receiver-operating characteristic curve (AUROC). For all analyses, two-sided statistical significance was defined as p <0.05. Data were analyzed using SPSS (statistical package for social science) version 25.0 (IBM SPSS Statistics for MAC OS. Armonk, NY: IBM Corp.)

Results
From an original cohort of 133 patients who met the inclusion criteria, 12 were lost to follow-up, 10 had unreliable LSM, and one was excluded for treatment failure. Therefore, 110 patients were included in the study, of which 48 (43.6%) were males, with a median age of 64 (52;74) years. Sixtythree patients (57.2%) had genotype 1 infection, 14 (Table 3). In particular, the difference in median SSM (45 vs. 19 kPa, p <0.001) and spleen area (54 vs. 38 cm 2 , p <0.001) were more marked between group 1 and group 4. Regarding portal vein flow velocity, patients in group 1 had slower values if compared to other groups, excluding group 1 and group 2 comparison. Inter-group differences of portal vein diameter were not statistically significant, except for group 1 and group 4 (1.14 vs. 1.01 cm, p = 0.043). Table 1. Patients' laboratory values prior to treatment, end of treatment, and one year after SVR 12. † represents a p <0.001, expressed as a comparison between baseline and one-year after SVR12; ‡ represents a p <0.001 expressed as a comparison between end of treatment and one-year after SVR12. AST: aspartate aminotransferase; ALT: alanine aminotransferase; GGT: gamma-glutamyl transpeptidase; ALP: alkaline phosphatase.

Values
Baseline End of Therapy

Models of Fibrosis Overestimation
We built four models according to logistic regression analysis, whose discriminative and calibration metrics are described in Table 4. In M1-AST and M1-ALT, the dependent variable is represented by the overestimation of one grade of fibrosis, whereas the independent variables are AST and ALT, respectively. The probability of overestimation of one grade of fibrosis is plotted in Figure 1. The two probability plots present different graphical behaviors: while the M1-ALT shape could be approximated to that of an S-shaped curve, M1-AST is almost linear. According to Table 4, M1-ALT showed better calibration (lowest AIC/BIC and highest Pseudo-R2) but equal discriminative ability. Also, according to these models, there is a probability to overestimate one grade of fibrosis >75% for ALT ≥150 IU/L (3.3 ULN) and AST ≥335 IU/L (7.4 ULN).  Table 4. Association between serum transaminase and liver fibrosis overestimation. Models were realized according to fibrosis measurement one-year after SVR12. VIF: variance inflation factor; AIC: Aikake information criterion; BIC: Bayesian information criterion; AUROC: area under the receiveroperating characteristic curve. In M2-AST and M2-ALT, the dependent variable is represented by the overestimation of two or more grades of fibrosis, whereas the independent variables are AST and ALT, respectively. The probability of overestimation of two or more grades of fibrosis is plotted in Figure 2. Although both curves are S-shaped, the M2-ALT slope increases more rapidly; from 0% at 50 IU/L to 99% at 127 IU/L. The two curves intersect at 84 IU/L, where the probability is equal to 46%. Also, M2-ALT reaches the superior plateau (equal to 100% of probability) at 140 IU/L, whereas M2-AST approaches it at 211 IU/L. In summary, the probability of overestimation of two fibrosis grade is equal to (1)

Discussion
It is commonly assumed that aminotransferase elevation is mainly related to liver cellular damage when hepatocytes undergo necrosis as a result of direct cellular damage or inflammation [24][25][26]. In the case of HCV, the cytolysis of infected hepatocytes is mediated by perforin and granzyme B produced by cytotoxic T-lymphocytes [27]. Also, serum transaminase concentrations (mostly ALT) have been linked to the histological grading of necroinflammatory activity, and, as expected, patients with elevated ALT generally show higher activity scores [28,29]. These concepts should be taken into consideration for a critical evaluation of liver elastography results as it has been proven that stiffness values are influenced by tissue congestion and inflammation, rather than the mere degree of collagen deposition (i.e., fibrosis) [30,31]. Therefore, it is safe to assume, as, in the case of our study, that hepatic congestion and inflammation can affect LSM, and that serum transaminase levels can be used to quantify the amount of ongoing damage. In this regard, even if their utility is still in debate [32], inflammation-adapted LSM cut-offs have been proposed by Mueller et al. [33]. In addition, successful viral eradication has been associated with a decline in LSM. According to a recent meta-analysis which included 24 studies, six to twelve months after achieving SVR12, patients experience a median LSM decrease of 28%, and approximately 47% of patients with baseline values compatible with advanced liver fibrosis or cirrhosis show post-treatment values compatible with lower degrees of fibrosis [34]. Also, in light of recent findings, the decrease of LSM appears to be more evident in patients with baseline ALT ≥ 2ULN [35]. In our cohort of patients, LSM showed a median decrease of 35% from baseline values, and 67 (61%) patients showed post-treatment values compatible with lower fibrosis stages. Notably, the patients who showed LSM compatible with F0-1 one year after SVR 12 (Table 2), were the ones with higher serum transaminases levels (median AST 3.9 ULN and median ALT 5.6 ULN), whereas those without fibrosis downstaging had lower median levels of transaminases (median AST 1.4 ULN and median ALT 1.6 ULN). These observations may suggest that the principal driver of LSM improvements is related to the suppression of virus-induced liver inflammation, as a consequence of successful viral eradication rather than a simple regression of liver fibrosis. This hypothesis is what drove us in trying to quantify the degree of overestimation related to serum transaminases: we developed M1s and M2s using as dependent variables the downstaging of one fibrosis stage (M1s) or two or more fibrosis stages (M2s) and serum transaminases as independent variables. It should be taken into account that the logistic regression models were built on the assumption that one year is a relatively short period for histological improvement of liver fibrosis, especially in patients who are supposed to have significant liver fibrosis [36][37][38]. Regarding Table 4, both M1s have significantly less performant calibration and discriminative metrics if compared to M2s. The slope of M2-ALT increases more rapidly from 50 IU/L to 127 IU/L, which means that variations in the probability of overestimation for the same unitary increment are peculiar to each transaminase. Also, M2-ALT showed slightly better calibration (lowest AIC/BIC and highest Pseudo-R2) and discriminative ability (highest AUROC) if compared to M2-AST. According to Figure 2, the probability of overestimation of two fibrosis grade is equal to 50% for serum transaminases around 2-2.2 ULN and steadily increase to 100% for serum transaminases between 3.1 and 4.7 ULN.
Alternatively, according to the methods section, these results can be interpreted as follows: what is the probability of LSM downstaging (≥2 fibrosis stage) at one year of follow-up, starting from values compatible with severe liver fibrosis, and considering baseline serum transaminase? As stated above, it is very unlike to obtain real downstaging in such a short follow-up period. However, it would be appropriate to detect accurate LSM after the elimination of such a robust confounding factor as inflammation. Other parameters can be used for appropriate liver fibrosis staging and, in particular, ultrasonographic signs of severe liver fibrosis. For example, SSM has been found to correlate with liver fibrosis staging [39,40]; besides, spleen size and platelet count are often used as noninvasive indicators of liver cirrhosis [41,42]. Prior to therapy, patients in G1 had significantly lower platelet counts if compared to all other groups ( Table 2). In addition, one year after SVR12, each fibrosis group was characterized by different ranges of SSM and spleen areas (Table 3). Although it could be challenging to differentiate between each fibrosis stage, it is possible to use SSM values to rule-in liver cirrhosis: G1 showed statistically higher SSM if compared to other groups.
The most important limitation of this study is the absence of the histological assessment of liver fibrosis, both at baseline evaluation and during follow-up, which is a common problem in the majority of the studies that are addressing this issue in the DAAs era. However, the real novelty of this study is related to the quantification of the influence of elevated serum transaminase levels in LSM, and the creation of a probability model that can be easily read and employed in the everyday clinical practice. We believe that our study may serve as a reference point for appropriate fibrosis stratification by liver elastography in patients with active HCV infection. Funding: The authors received no specific funding for this work.

Conflicts of Interest:
The authors declare no conflict of interest.